Skip to content

A free academy for AI and agents

Understand AI, then build agents.

Short chapters, a visual you can play with, and a check. Everything runs in your browser. No account, no sign-up.

  • Runs in your browser
  • No account
  • Free

Learning paths

Start with the ideas, or start with code.

The first track explains AI, models and agents, with no code. The second is for developers who build agents. In every path, each chapter builds on the one before.

Understand AI

How AI and its models work, what agents are, and how to use them well. No code.

  1. 018 chapters
    BeginnerNo code

    AI principles

    What AI is, how it writes, where it fails, and how to use it with care. No code.

    Before you start
    Nothing.
  2. 028 chapters
    BeginnerNo code

    Understanding models

    The kinds of models agents use, where they come from and how to compare them. Numbers, no code.

    Before you start
    Nothing. The first two chapters of AI principles help.
  3. 038 chapters
    BeginnerNo code

    Agents for everyone

    What AI agents are, how they work and how to use them well. No code.

    Before you start
    Nothing. If you have used a chatbot, you are ready.

Build agents

For developers: a first agent in LangGraph.js, then systems of agents, then long and complex runs.

  1. 018 chapters
    BeginnerFor developers

    Build your first agents

    From one model call to a guarded agent, with LangChain.js and LangGraph.js, in your browser.

    Before you start
    Some JavaScript: functions, objects, async/await and installing packages with npm.
  2. 028 chapters
    IntermediateFor developers

    Agent systems and standards

    MCP, A2A, AG-UI and friends, orchestration, context, tracing and testing, with the real protocol code.

    Before you start
    The Build your first agents path, or the same knowledge: LangChain.js messages and tools, LangGraph graphs and the agent loop.
  3. 0310 chapters
    AdvancedFor developers

    Advanced agents

    State at scale, Command and Send, subgraphs, time travel, loops that survive and loops that stop, planning, deep agents, and jobs that span sessions.

    Before you start
    The Agent systems path, or the same knowledge: graphs, tools, the agent loop, checkpoints and traces.

How it works

Every chapter works the same way.

  1. Read

    A short text, 400 words at most, that explains one idea and every term it uses.

  2. Try the visual

    An interactive picture of the idea. Do the thing it asks and the goal turns green.

  3. Pass the check

    A question, an order to fix or an output to predict. You can retry as often as you like.

  4. Keep your place

    Progress stays in this browser. Export it to a file to carry on in another one.

What is real

We say what runs for real and what is simulated.

Every visual carries one line that tells you which parts are real, re-created or simulated.

Real

Code that runs in this page: small models trained as you use them, LangGraph.js graphs, the official MCP SDK and the protocol messages you see on screen. Output recorded beforehand from real tokenizers and models is labelled as recorded.

Re-created

Behaviour copied from software that cannot run in a browser and rebuilt to act the same way, or answers recorded beforehand from a model.

Simulated

The model's decisions are scripted, so you need no key and the run is the same every time. Data in examples is fictional.